A simple Kohya ss LoRA (not multi-trained, as my recent models have been) of the Australian actress and latterly medical practitioner <a target="_blank" rel="ugc" href="https://en.wikipedia.org/wiki/Pamela_Stephenson">Pamela Stephenson</a>, who shot to fame in the UK at the end of the 1970s in the satirical show <a target="_blank" rel="ugc" href="https://en.wikipedia.org/wiki/Not_the_Nine_O%27Clock_News">Not The Nine O'Clock News</a>, but was later introduced to a wider audience through a <a target="_blank" rel="ugc" href="https://www.youtube.com/watch?v=Ig5uMkprqEk">prominent appearance</a> in Superman III (1983).This is probably the worst resemblance of any of my models, and this is the fifth time I have attempted a LoRA of this person. In any case, I don't think I can do any better under the circumstances. Incredibly, <a target="_blank" rel="ugc" href="https://www.reddit.com/r/StableDiffusion/comments/16tywgr/tutorial_new_workflow_for_15_lora_training/">multi-training</a> seems to actually make it worse!There are a few big issues with doing a LoRA of Pamela Stephenson: one is that she was so perfect-looking in the 1978-83 period, that she already looks like a Stable Diffusion generation for 'attractive blonde woman'. It's the imperfections and deviations from the standard that gives people character, and more particular beauty.The second is that there is very little suitable material out there, considering how fondly-remembered she is from this period.Thirdly, if you drag any image that this LoRA makes into an image-based search on Yandex or at <a target="_blank" rel="ugc" href="https://clarifai.com/clarifai/main/models/celebrity-face-recognition?tab=overview&amp;inputId=https%3A%2F%2Fs3.amazonaws.com%2Fsamples.clarifai.com%2Ffeatured-models%2Fcelebrity-kim-kardashian.jpg">Clarifai's celeb detector</a>, the image will almost certainly be flagged as Kim Basinger, or the equally android-like Cheryl Ladd, both of whom are better-represented, both in search algorithms and in the LAION database powering Stable Diffusion.For the latter reason, it is very hard to get this LoRA to stop 'Basinger-ing'/'Ladd-ing' renders! If you turn the LoRA strength above 1, it can help to let Stephenson's features break through.(However, if you drag the same image into the far better <a target="_blank" rel="ugc" href="https://pimeyes.com/en">PimEyes engine</a>, it will identify the render as Pamela Stephenson)This LoRA may produce NSFW features - I wouldn't know. The image that accompanies this post is the only one I have produced from it so far.One thing is certain - if you want a period-accurate likeness with this LoRA, you're going to have to put in extra time!Use a VERY low CFG for this (4, 5 etc).

Pamela Stephenson (1979-83 era) SD1.5 LoRA

This is an ordinarily trained LoRA and does well around the 1.0 LoRA strength mark. Keep CFG very low.

Files with SHA256: 6f57919e9fb9bae4aa2243e01e58912e0194ac50819a4fca4933e661807f5352